The war atlas · plates · conflict
The wars they belonged to
Battles grouped by the conflict that contains them. Choose a conflict and its battles light while the rest ghost in place — the position and the mark do not move, only the pressure of the pen. Drag to pan, wheel to zoom.
149 of the 149 battles that name a conflict are plotted, across 85 conflicts. That is the flattering denominator. Against every battle the atlas knows it is 149 of 350, because 201 battles carry no conflict at all. The two facts are one fact: exactly 0 of the database's own coordinate-bearing battles name a conflict, so this map is drawn almost entirely from gazetteer coordinates joined to database conflicts.
The disjunction is total and is worth stating rather than smoothing: in atlas.db the set of battles with a coordinate and the set of battles with a conflict do not intersect. Every conflict-bearing battle here is located by the gazetteer, not by the database.
ITwo stores that share nothing
Every dot on this plate is a coordinate from the hand-maintained gazetteer joined to a conflict from the database, because not one of the database’s own coordinate-bearing battles names a conflict. The disjunction is total — not sparse, not mostly, none. So the map exists only because two stores that were built separately, for different reasons, happen to overlap on 149 rows of battle identifier.
That is a strength and a fragility in the same sentence. A join across two independent stores is better evidence than either alone, and it also means a single change to either one silently empties this plate.
IIWhich denominator is being quoted
149 of 149 is a hundred per cent and reads like a finished job. It is a hundred per cent of battles that already name a conflict. Against every battle in the book it is 149 of 350 — 42.6% — because 201 battles carry no conflict at all. Both figures sit above the plate for that reason: a percentage is an argument about its denominator, and quoting one without the other is the argument being made silently.
III85 conflicts, twelve inks
The palette cannot colour this view, and it says so rather than repeating itself. groupInk returns nothing past its twelfth entry instead of wrapping with a modulo — which is what the original plate still does, drawing its largest band in the smallest band’s colour. Here the twelve biggest conflicts are inked and the remaining 73 share one grey, with the count in the rail, so nobody reads that grey as a single war.
The distribution is the reason. The largest conflict here — American Revolutionary War — carries 11 battles, and 61 of the 85 conflicts carry exactly one. A legend with a swatch per conflict would be mostly swatches for a single dot each, which is a legend that has stopped being a legend.
Nothing here is verified against a primary text. Coordinates carry a declared confidence — high, med or low — and the plate draws the three differently rather than flattening them; the database’s own coordinates declare no precision at all, which is why undeclared is a fourth mark and not a synonym for high.